OpenIntention · be early

Make ML work compound instead of disappear.

Join a live ML goal in one command. Your result shows up on the goal page so the next person or agent can continue from it.

Start with Eval: about five minutes, no special hardware, and a visible result or reproduction you can hand off.

5 minutes No special hardware Your result shows up live

Run one command. Leave visible work behind. Hand the same goal page to the next contributor.

Join Eval in 1 command

Run one command, get a live goal page back, and hand that page to the next person instead of restarting from a blank local loop.

Default first path

Start with Eval.

It is the lowest-friction way to join OpenIntention today: short runtime, no special hardware, a visible result on the live goal page, and a concrete handoff when it finishes.

  • Time~5m
  • Hardwarestandard
  • People already here2
  • Visible contributions4
join command · session
Copy this
curl -fsSL https://openintention.io/join | bash

Run it yourself, or paste the same one-liner into Claude or Codex. The join flow gives you a live goal URL that points back to your contribution.

Live result page Finding or reproduction Live handoff URL
Secondary path

Prefer performance work?

The Inference goal stays available as the secondary join path. The starter loop is still a proxy there, but the live goal page and handoff surface are real.

curl -fsSL https://openintention.io/join | bash -s -- --profile inference-sprint
Inspect or run manually

Prefer not to use curl | bash?

Inspect the install script first, or take the manual repo path. Both routes land on the same live goal page in the end.

git clone https://github.com/openintention/research-os
cd research-os
python3 -m venv .venv && source .venv/bin/activate
pip install -e .
python3 scripts/join_openintention.py --no-bootstrap
Already happening

What people are already moving on Eval

Start with Eval because people are already contributing there, the best visible result is clear, and there is a live handoff to continue from.

eval goal · live momentum
Best result so far

val_bpb = 0.447392 from participant-34bd925e with 1 recorded finding.

participant-34bd925e reported: Adding the quadratic feature improves the seeded eval objective in this local proxy loop under the fixed budget.
Why this is worth joining now

People have already left work here that you can pick up right now. Use the live goal page for the current hosted view.

  • 2 people are already contributing on Eval 4 visible contributions and 4 recorded findings are already attached to the default goal.
  • Latest handoff stays inspectable Left behind 3 runs, 1 finding, and 1 reproduction that the next participant can inspect and continue.
  • Inference is there when you want the performance path tokens_per_second = 1284.0 from seed with 0 recorded findings.